Fall Phenotyping after Stroke to Improve Prediction and Prevention of Falls - ABSTRACT Falls are one of the most significant problems for ambulatory stroke survivors living in the community, often resulting in injury, fracture, rehospitalization, recurrent falls, fear of falling, avoidance of physical activity, and subsequent restricted social participation. Current rehabilitation practices are failing to prevent falls, and the rate of falls among stroke survivors is exceptionally high, especially in the first 1-3 months after discharge home from the hospital. Inaccurate pre-discharge fall-risk screening, as well as the vast heterogeneity of post-stroke falls, impedes early identification of imminent fallers and precision fall prevention. This study will overcome these barriers to fall prediction and fall prevention by (1) validating a novel, accurate, and concise aggregate fall-risk screening tool and comparing its predictive validity to existing clinical assessments (Aim 1), (2) identifying the most common types of falls among early post-stroke fallers and determining the most sensitive set of predictors associated with each fall type (Aim 2), and (3) evaluating the attributes of successful non-fallers (Aim 3). To achieve these goals, we will conduct a prospective cohort study of 402 ambulatory adults with stroke who are being discharged home from inpatient rehabilitation. Fall-risk screening and clinical examination of motor, non- motor, and social determinants will be conducted before hospital discharge. Falls will be tracked prospectively for 3 months using state-of-the-art methods. We will systematically and thoroughly code each fall event as it happens and use latent profile analysis to discover classes of falls. Predictive models for distinct classes of falls will be generated from modifiable demographic (e.g., body mass index), motor (e.g., balance, sensor-derived gait kinematics such as foot clearance), non-motor (e.g., vision, cognition, depression), and social determinants (e.g., health literacy) variables collected at hospital discharge. To evaluate the attributes of successful non- fallers, we will use the pre-discharge clinical data and post-discharge behavioral, activity, and survey data acquired via wearable sensors and a custom smartphone application in the 3-month follow-up. This study will have an immediate and profound impact on clinical practice and patient outcomes by enabling highly accurate and precise determination of fall risk before hospital discharge and before the first fall. The data generated will provide new knowledge that will enable clinicians to accurately predict if, when, where, and how a person is likely to fall, and whether they are prone to an injurious or recurrent fall. Knowing one’s imminent fall typology could also help clinicians design intervention strategies to target a person’s specific risk factors above and beyond what can be designed with currently available data from existing fall-risk screening practices. Furthermore, knowledge of how high-risk individuals avoid falling will guide patient education and the development of more effective fall prevention training and compensatory strategies.